[BioSQL-l] PhyloDB module updates

[Hilmar Lapp]

On Feb 14, 2008, at 10:32 AM, aaron.j.mackey at gsk.com wrote:

>>> Though speaking of which, what has been bugging me lately is that
>>> they only provide for a single taxonomy to be used (presumably the
>>> one from NCBI, since one column expressly refer to the
>>> NCBI_taxon_ID. I've been thinking about suggesting to generalize
>>> that a little, but the backwards compatibility issues might just
>>> outweigh the benefits.
>
> One approach would be to reuse the existing ontology framework (and
> perhaps include a NCBI-taxonomy-specific "view" to recapitulate the
> existing taxon tables).  This would have the added advantage of  
> allowing
> DAG-like taxonomies.

Indeed that's a strong possibility too, and in fact for example the  
NCBI taxonomy can be downloaded from the OBO foundry in OBO format  
(and also in OWL I believe). We are also converting another taxonomy  
(Eschmayer's Catalog of Fishes) into an ontology (Teleost Taxonomy  
Ontology - TTO - can also be downloaded from the NCBO).

That said, most taxonomies aren't (yet?) available as an ontology,  
and for the exercise of converting there are a couple of gotchas that  
require manual review of the result.

Maybe the ideal level of support for this (i.e., when a taxonomy  
doesn't come as an ontology to begin with) would be in the toolkit 
(s), mapping from a taxonomy model (Bio::Taxonomy, or Bio::Tree, for  
example?) to the ontology model (Bio::Ontology).

(This is mostly a stream of thoughts, not clear arguments one way or  
another.)

> "taxon_id" foreign keys would then instead be
> "term_id" (too bad there's no such thing as column-level aliases/ 
> views).

Indeed. But there's more than that needed. For example, ranks would  
be extra relationships ('has_rank') that would need to be joined in.

	-hilmar

-- 
===========================================================
: Hilmar Lapp  -:-  Durham, NC  -:-  hlapp at gmx dot net :
===========================================================